Crowdsource the Metadata (Stop Being the Bottleneck)
The Problem
Your team maintains a "data catalog." Except it's always wrong. The data engineer who built Table_X left 2 years ago and nobody knows what field_47b means. You have 500 tables and 3 people trying to document them. It's 18 months out of date. Users don't trust it. So they ask you directly, 50 times a week: "What's this field?" You're a human dictionary. Waste. And you can't scale—if documentation requires YOU, you'll never keep up with 500 tables growing to 5,000.
The Principle
Let users document what they know. Wikipedia works because millions contribute, not because one team writes everything. Same with your data. Crowdsource metadata: what tables mean, which fields matter, data quality issues, example queries. Users who USE the data write the docs. Your job? Curate, not create. Set up a system where anyone can add descriptions, tag tables, rate quality, leave examples. The magic happens when 100 users each document 5 tables instead of 3 people trying to document 500.
3 Steps to Take Action
- 1
**Enable User Contributions** Add to your data catalog: - 'Add description' button (any user can contribute) - Star ratings for data quality - Comment threads for clarification - 'Example queries' section - Tags/labels Make it stupidly easy to contribute. Think Wikipedia, not a 47-field form. Test with 5 power users: 'Document your favorite 3 tables.' If it takes >5 minutes per table, simplify.
- 2
**Gamify Contributions** Create incentives: - Leaderboard: most contributions this month - Badges: 'Data Documentor', '10-Table Club' - Recognition: shout-outs in team meetings - Make it visible: 'Last edited by Sarah' stamps Don't require perfection. Accept 'good enough' contributions. You can curate later. Goal: 50% of users contribute at least one piece of metadata in 90 days.
- 3
**Curate, Don't Create** Your team's new role: - Review user contributions weekly - Fix obviously wrong info - Merge duplicate entries - Promote great examples to 'featured' - Fill gaps where no one contributes Track: % of tables with user-contributed metadata. Target: 70%+ within 6 months. If a table has zero contributions after 90 days, either nobody uses it (delete it) or you document it yourself.
💡 Inspired By
Wikipedia, StackOverflow, and A major bank's crowdsourced BI catalog. The wisdom of crowds beats the bottleneck of experts.
This is what Analytics Pitfalls & Principles is about: people helping people.
In Practice
A financial services data team had 847 tables with 14% documented (120 tables). Metadata quality was so bad users bypassed the catalog entirely. They tried hiring more documenters—still couldn't keep up. Pivoted: built a Wiki-style data catalog where ANY user could add descriptions, tags, quality ratings, example SQL. Gamified it: leaderboard for most contributions. Within 6 months: 673 tables documented (79%). Users trusted it because it was written by users who actually used the data. Data team shifted from writing docs to moderating and curating. Governance through crowdsourcing. Questions dropped 68% because users found answers in the catalog.
